White House Tightens Control Over Frontier AI, Partially Lifts Ban on One Anthropic Model

Overview

• In early June the president signed an executive order proposing a voluntary 30-day vetting process for advanced models, but agencies moved to stronger tools before that framework was implemented.
• In mid-June Commerce and security agencies imposed export controls tied to reported jailbreak and model-extraction vulnerabilities, forcing Anthropic to disable some models globally and opening daily technical negotiations with agencies.
• Late in June the administration partially rescinded the ban on one Anthropic model while keeping Fable 5 blocked, and it pressed OpenAI to restrict GPT-5.6 access to a small, administration-approved group as talks continue.
• AI companies, trade groups and former advisers are demanding a clear, repeatable licensing or vetting process because they say ad hoc, one-off decisions are creating regulatory unpredictability and slowing releases.
• The policy shift has slowed product rollouts, raised legal and diplomatic questions about using export rules for cloud and API access, and risks handing an edge to foreign labs if U.S. firms sit in regulatory limbo.